• Mandarin Chinese has a five-vowel system

1. [yi], [yu], [wu], [en] and [ai]
2. Four allophones for the mid vovel: [e], [o], [en] and [e]
3. Two allophones for the low vovel: [ai] and [ao]

• What are the individual variances when producing these 9 vowels?
• Vowel space in Chinese?

This study investigates native Chinese speakers' vowel distribution and non-native Chinese speakers' vowel distribution

vocal tract

Participants

• 6 female native Mandarin speakers (NS) & 6 female non-native Mandarin speakers (NNS)
• Average age

1. NS Mandarin speakers was 26.3 (range: 23-30)
2. NNS Mandarin speakers was 22.3 (range: 18-28)

• Length of study: 6 months

Speech materials

• 9 Chinese sentences
• Each sentence includes one of the following nine vowels: [yi], [yu], [wu], [e], [o], [en], [e], [ai], [ao].
• Each vowel appears after the aspirated bilabial stop [p] with a high tone (55).

Procedure

• Sentence-repetition oral task
• Non-native speakers and native Chinese speakers were asked to read the sentences twice.
• Formant 1 (F1) and Formant 2 (F2) were measured in the vowel mid-point for each vowel shown in the spectrogram through Praat.

Google @datalorax

## # A tibble: 6 x 14
##   X__1  gender height   age Mean     yi    yu    wu    ye    wo    en     e
##   <chr> <chr>   <dbl> <dbl> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 NS1   F         163    29 F1      270   326   320   583   553   730   634
## 2 NS1   F         163    29 F2     3041  2678   927  2548   872  1803  1078
## 3 NS2   F         164    30 F1      321   307   341   537   563   710   482
## 4 NS2   F         164    30 F2     2909  2383   836  2347   958  1863  1201
## 5 NS3   F         165    27 F1      300   316   321   547   472   740   553
## 6 NS3   F         165    27 F2     2899  2481   776  2694   821  1812  1282
## # ... with 2 more variables: ai <dbl>, ao <dbl>

• Data is definitely NOT tidy!
  • Variables (F1, F2) don't have thier own columns.
  • Each row includes multiple observations from multiple vowels.
• Confidence level: 5/5 with dplyr::spread and dplyr::gather

names(df)

##  [1] "X__1"   "gender" "height" "age"    "Mean"   "yi"     "yu"    
##  [8] "wu"     "ye"     "wo"     "en"     "e"      "ai"     "ao"

• Varibles' name are not clean.
  • Confidence level: 5/5 with janitor::clean_names

print(df$X__1)

##  [1] "NS1"  "NS1"  "NS2"  "NS2"  "NS3"  "NS3"  "NS4"  "NS4"  "NS5"  "NS5" 
## [11] "NS6"  "NS6"  "NNS1" "NNS1" "NNS2" "NNS2" "NNS3" "NNS3" "NNS4" "NNS4"
## [21] "NNS5" "NNS5" "NNS6" "NNS6"

• X__1 contains id and group (NS & NNS).
  • Confidence level: 2/5 with dplyr::separate
    • sep = ?
• Duplicate id number for NS and NNS
  • Confidence level: 0/5
    • use dplyr::case_when ?

tidy_df1 <- df %>% 
  clean_names(case = "snake") %>% 
  separate(x_1, into = c("group", "id"),
           sep = "(?<=[A-Za-z])(?=[0-9])") %>%  #sep = -1 also works
  mutate(id = case_when(
   group == "NNS" & id == '1' ~  '7',
   group == "NNS" & id == '2' ~  '8',
   group == "NNS" & id == '3' ~  '9',
   group == "NNS" & id == '4' ~  '10',
   group == "NNS" & id == '5' ~  '11',
   group == "NNS" & id == '6' ~  '12',
   TRUE ~ id
   )) %>% 
  mutate(group = as.factor(group),
        id = as.numeric(id))   

tidy_df2 <- tidy_df1 %>% 
  gather(vowel, value, -1:-6) %>%
  mutate(vowel = as.factor(vowel)) 

head(tidy_df2)

## # A tibble: 6 x 8
##   group    id gender height   age mean  vowel value
##   <fct> <dbl> <chr>   <dbl> <dbl> <chr> <fct> <dbl>
## 1 NS        1 F         163    29 F1    yi      270
## 2 NS        1 F         163    29 F2    yi     3041
## 3 NS        2 F         164    30 F1    yi      321
## 4 NS        2 F         164    30 F2    yi     2909
## 5 NS        3 F         165    27 F1    yi      300
## 6 NS        3 F         165    27 F2    yi     2899

tidy_df <- tidy_df2 %>% 
  spread(mean, value) %>%
  #reorder variables & discard gender
  select(id, group, age, height, vowel, F1, F2) %>% 
  arrange(id)

head(tidy_df)

## # A tibble: 6 x 7
##      id group   age height vowel    F1    F2
##   <dbl> <fct> <dbl>  <dbl> <fct> <dbl> <dbl>
## 1     1 NS       29    163 ai      808  1802
## 2     1 NS       29    163 ao      832  1531
## 3     1 NS       29    163 e       634  1078
## 4     1 NS       29    163 en      730  1803
## 5     1 NS       29    163 wo      553   872
## 6     1 NS       29    163 wu      320   927

figure1_pre <- tidy_df %>% 
  ggplot(aes(x = F2, y = F1, color = vowel)) + 
  geom_point(size = 3) +
  facet_wrap(~ group) + 
  theme_classic() 

• Couple issues:
  • X and Y axes need to be rotated.
  • Labels are not in a conventional order.

figure1_post <- tidy_df %>% 
  ggplot(aes(x = F2, y = F1, color = vowel)) + 
  geom_point(size = 3) +
  facet_wrap(~ group) + 
  theme_classic() +
  
  #reverse x and y to meet the perception of sounds
  scale_x_reverse() +  
  scale_y_reverse() +
  
  #reorder vowel based on IPA order
  scale_color_discrete(breaks = c("yi", "yu", 
                                  "wu", "ye", 
                                  "wo", "en", 
                                  "e", "ai", 
                                  "ao")) +
  labs(title = "Vowel Distribution among All Non-native and Native Speakers")

• Hard to see group differences.
• Inefficient references due to 9 vowels.

• Google
• Special Thank You to Joey Stanley !
  • Joey's blog
  • Making vowel plots in R

smry_df <- tidy_df %>% 
  group_by(group, vowel) %>% 
  summarize_at(vars(F1, F2), funs(mean, sd)) 
  

figure2 <- smry_df %>% 
  ggplot(aes(x = F2_mean, y = F1_mean)) +
    
    # adds text directly to the plot
    geom_label(aes(label = vowel, color = group)) +
  
    scale_x_reverse() +  
    scale_y_reverse() +
    theme_classic() +
    labs(title = "Mean Vowel Distribution between Non-native and Native Speakers",
         x = "F2",
         y = "F1")

figure3 <- 
  ggplot(data = tidy_df, aes(x = F2, y = F1 )) +
  geom_label(data = smry_df, aes(x = F2_mean, y = F1_mean, 
                                 label = vowel, fill = group),
             alpha = 0.2) +
  geom_point(aes(color = vowel, shape = group),
             size = 3, alpha = 0.4) +
  stat_ellipse(aes(color = vowel), level = 0.67) +
  scale_x_reverse() +  
  scale_y_reverse() +
  theme_classic() +
  labs(title = "Individual and Group Vowel Distribution",
       subtitle = "The size of output ellipses in standard deviations.") +
  guides(color = FALSE)

smry_df %>% 
  kable(format = "html",digits = 2,caption = "Table1",booktabs = T)

kable(smry_df)

table_f1mean <- smry_df %>% 
  select(1:3) %>% 
  spread(vowel,F1_mean) 

table_f1mean

## # A tibble: 2 x 10
## # Groups:   group [2]
##   group    ai    ao     e    en    wo    wu    ye    yi    yu
##   <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 NNS    913.  902.  640.  651.  552.   416  521.  336.  322.
## 2 NS     910.  848   597.  718.  554.   335  557.  308.  309.

table_f2mean <- smry_df %>%
  select(1:2,4) %>% 
  spread(vowel,F2_mean) 

table_f2mean

## # A tibble: 2 x 10
## # Groups:   group [2]
##   group    ai    ao     e    en    wo    wu    ye    yi    yu
##   <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 NNS   1514. 1377  1702. 1980  1044. 1123. 2320  2646. 1807.
## 2 NS    1656. 1305. 1290. 1850.  909.  840  2486. 2916. 2421.

bind_rows(table_f1mean,table_f2mean) -> bindtables 
bindtables

## # A tibble: 4 x 10
## # Groups:   group [2]
##   group    ai    ao     e    en    wo    wu    ye    yi    yu
##   <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 NNS    913.  902.  640.  651.  552.  416   521.  336.  322.
## 2 NS     910.  848   597.  718.  554.  335   557.  308.  309.
## 3 NNS   1514. 1377  1702. 1980  1044. 1123. 2320  2646. 1807.
## 4 NS    1656. 1305. 1290. 1850.  909.  840  2486. 2916. 2421.

bindtables["Mean"] <- c("F1","","F2","")
bindtables

## # A tibble: 4 x 11
## # Groups:   group [2]
##   group    ai    ao     e    en    wo    wu    ye    yi    yu Mean 
##   <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>
## 1 NNS    913.  902.  640.  651.  552.  416   521.  336.  322. F1   
## 2 NS     910.  848   597.  718.  554.  335   557.  308.  309. ""   
## 3 NNS   1514. 1377  1702. 1980  1044. 1123. 2320  2646. 1807. F2   
## 4 NS    1656. 1305. 1290. 1850.  909.  840  2486. 2916. 2421. ""

bindtables_new <- bindtables[colnames(bindtables)[c(11,1:10)]]
bindtables_new

## # A tibble: 4 x 11
## # Groups:   group [2]
##   Mean  group    ai    ao     e    en    wo    wu    ye    yi    yu
##   <chr> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 F1    NNS    913.  902.  640.  651.  552.  416   521.  336.  322.
## 2 ""    NS     910.  848   597.  718.  554.  335   557.  308.  309.
## 3 F2    NNS   1514. 1377  1702. 1980  1044. 1123. 2320  2646. 1807.
## 4 ""    NS    1656. 1305. 1290. 1850.  909.  840  2486. 2916. 2421.

bindtables_new %>% 
  kable(format = "html",digits = 2,caption = "Table 1", booktabs = T) 

An ideal table

• The importance of IPA symbols
  • How to SHOW IPA symbols in RStudio?
  • How to KNIT IPA symbols to PDF in APA format?

install.packages("linguisticsdown")

## Error in contrib.url(repos, "source"): trying to use CRAN without setting a mirror

devtools::install_github("liao961120/linguisticsdown")

Write IPA symbols with Phonetic Features

• output:
• papaja::apa6_pdf:
• toc: true
• toc_depth: 5
• latex_engine: xelatex
• includes:
• in_header: header.tex

• header-includes:
• 
  

\newfontfamily\ipa{Doulos SIL} 

• IPA symbols cannot be properly rendered when output to PDF.

IPA symbols in R Markdown

• 3 out of 9 symbols disappeared in plots.

• How to support Linguistics-related doucument writing in R Markdown?
  • linguisticsdown
  • phonTools
  • ….

• Conditional compilation
• Templates: APA article & Support_IPA
• In-line codes for plots?
• r linguisticsdown::cond_cmpl("ə")